vv

app.py

@@ -42,16 +42,14 @@ def segment(image_pil: Image.Image, prompt: str):
     # Convert PIL image to numpy array
     image_np = np.array(image_pil)
 
-    # Ensure the model is explicitly set to CPU
-    grounder.to("cpu")
-
     # Run GroundingDINO to get boxes for the prompt
     boxes, _, _ = predict(
         model=grounder,
-        image=image_np,
+        image=torch.tensor(image_np).to(device),  # Ensure image is on CPU
         caption=prompt,
         box_threshold=0.3,
-        text_threshold=0.25
+        text_threshold=0.25,
+        device="cpu"  # Explicitly set device to CPU
     )
     if boxes.size == 0:
         raise ValueError("No boxes found for prompt.")
@@ -97,3 +95,5 @@ if __name__ == "__main__":
     parser.add_argument("--port", default=7860, type=int)
     args = parser.parse_args()
     app.run(host=args.host, port=args.port)
+    args = parser.parse_args()
+    app.run(host=args.host, port=args.port)